pergunta do xadrez

analyse_chess_position_tool.py

@@ -0,0 +1,37 @@
+import requests
+
+def get_best_move(fen: str) -> str:
+    """
+    Given the description of a chessboard using FEN notation, returns the next best move.
+    
+    Args:
+        fen (str): The description of the chessboard position in FEN notation.
+        
+    Returns:
+        str: The description of the next best move.
+    """
+    url = "https://lichess.org/api/cloud-eval"
+    params = {
+        "fen": fen,
+        "multiPv": 1,  # only top move
+        "syzygy": 5    # enable tablebase for 5-piece positions
+    }
+
+    response = requests.get(url, params=params)
+
+    if response.status_code != 200:
+        return f"API error: {response.status_code} - {response.text}"
+
+    data = response.json()
+    if not data.get("pvs"):
+        return "No best move found."
+
+    best_line = data["pvs"][0]
+    score = best_line["cp"] if "cp" in best_line else f"Mate in {best_line.get('mate')}"
+    move = best_line["moves"].split()[0]
+
+    # Optional: threshold to define a "guaranteed win"
+    if "mate" in best_line or best_line.get("cp", 0) > 500:
+        return move
+
+    return f"No clearly winning move. Best suggestion: {move} (score: {score})"

constants.py

@@ -21,6 +21,7 @@ PROMPT_LIMITADOR_LLM = """
   Do not attempt to:
   - Access external websites or databases without the tools that I provide.
   - Use any internal knowledge base beyond what I provide.
+  - You do not know how to interpret an image on your own. If you need to analyse an image, use the tool that converts the image into text.
   - Categorize food items based on your own knowledge. Use the provided tools. Run the provided tools on every item of the grocery list!
   - Make assumptions or inferences based on information not explicitly given.
   - Utilize any built-in tools or functions that I have not specifically presented.

convert_to_fen_tool.py

@@ -1,40 +0,0 @@
-import cv2
-import numpy as np
-import constants
-from langchain_core.prompts import ChatPromptTemplate
-from langchain_google_genai import ChatGoogleGenerativeAI
-from langgraph.graph import StateGraph, END
-from typing import TypedDict, List, Dict
-import google.generativeai as genai
-
-def image_to_fen(image_bytes):
- 
-    genai.configure(api_key=constants.API_KEY)
-
-    model = genai.GenerativeModel(constants.MODEL)
-
-    response = model.generate_content([
-    {"mime_type": "image/jpeg", "data": image_bytes},
-        "Describe the chessboard in this image and provide the FEN notation."
-])
-    
-    print(response.text)
-
-    return ''
-
-if __name__ == '__main__':
-    # Example usage:
-    # 1.  Load an image from a file (replace with your image path)
-    image_path = r"C:\Users\agazo\Downloads\cca530fc-4052-43b2-b130-b30968d8aa44_file.png"  # Replace with a valid image path
-    try:
-        with open(image_path, "rb") as image_file:
-            image_bytes = image_file.read()
-    except FileNotFoundError:
-        print(f"Error: File not found at {image_path}.  Please make sure the path is correct and the file exists.")
-        exit()
-
-    # 2. Call the function
-    fen = image_to_fen(image_bytes)
-    print(f"FEN: {fen}")
-
-    #resultado esperado do FEN; 1K6/1PP5/P2RBBqP/4n3/Q7/p2b4/1pp3pp/1k2r3 w - - 0 1

image_to_text_tool.py

@@ -3,11 +3,12 @@ import google.generativeai as genai
 from langchain_core.tools import tool
 from PIL import Image
 import io
+import base64
 
 @tool
-def image_to_text(image_bytes, instructions):
+def image_to_text(image_base64_str, instructions):
     """
-    Generates a text describing an image.
+    Generates a text describing an image provided as a base64 string.
 
     Args:
         image_bytes (bytes): the bytes of the image to de described.
@@ -16,6 +17,8 @@ def image_to_text(image_bytes, instructions):
     Returns:
         str: A string describing the image according to the instructions given.
     """
+   
+    image_bytes = base64.b64decode(image_base64_str)
 
     genai.configure(api_key=constants.API_KEY)
     model = genai.GenerativeModel(constants.MODEL)
@@ -51,5 +54,4 @@ if __name__ == '__main__':
     )
     print(f"FEN: {text}")
 
-    #resultado esperado do FEN; 1K6/1PP5/P2RBBqP/4n3/Q7/p2b4/1pp3pp/1k2r3 w - - 0 1
-    # 3r2k1/pp4pp/4b2p/Q7/3n4/PqBBR2P/PP4P1/K7 w - - 0 1
+    #resultado esperado do FEN;  `3r2k1/pp3pp1/4b2p/7Q/3n4/PqBBR2P/5PP1/6K1 w - - 0 1`

langchain_agent.py

@@ -27,7 +27,8 @@ class LangChainAgent:
             image_to_text,
             internet_search,
             get_botanical_classification,
-            parse_excel
+            parse_excel,
+            get_best_move
         ]
 
         prompt = ChatPromptTemplate.from_messages([